INSIDECLAUDE DESKTOP
Secured by Mavs
A secure gateway that lets your power users run Claude Desktop inside company policy, with every action logged as evidence and conversation history that stays on their machine.
Runtime Security ForYour Most Advanced Users
Secure sensitive data at runtime within Claude Desktop
Values in prompts and files are replaced with synthetic stand-ins before the model sees them. Context intact.
Single Sign-On and Fleet Deployment
Push the client through MDM, and people sign in through your own identity provider (e.g. Microsoft Entra ID) rather than individual Anthropic accounts.
Storage & Data Sovereignty
Standard Claude Desktop keeps conversation history on Anthropic's servers. With the Mavs gateway it never leaves the user's local disk.
Configure connectors for each team
Decide centrally which connectors each team gets, instead of leaving it to per-task approval.
Guaranteed Data Residency
Inference runs only in the regions you nominate, with no traffic to claude.ai. That answers the residency question GDPR and sovereign data rules ask.
Consumption-Based Pricing
Standard deployment relies on seat licenses or flat-rate subscriptions. With Mavs you move to pure token-based billing, so you can roll out Claude Desktop to more people.
Frequently Asked Questions
What does Mavs add to Claude Desktop?
Claude Desktop gives your team an agent that reaches real files and real connectors and then acts on them. Mavs sits in front of it as the gateway every request routes through. Sensitive values in prompts and files are replaced before the model sees them, policy is enforced per team, and every action lands in a log you can hand to an auditor.
Doesn't Anthropic's Compliance API already cover this?
Claude Desktop is not covered by it, so whatever your team does inside Claude Desktop sits outside that reporting. Mavs covers it directly.
We already export OpenTelemetry. Isn't that our audit trail?
OpenTelemetry is telemetry, not audit logging. It tells you what happened operationally: which calls ran, how long they took, what errored. It does not give you an immutable record of every sensitive value that was replaced, which is what an auditor asks to see.
What does the model actually receive?
Synthetic stand-ins in place of the sensitive values, with the surrounding context intact. The agent can still do the work because the substitutes behave like the real values. On the way back, the real values are restored in place of the stand-ins, so what your employee sees is an ordinary Claude Desktop session. The real data never reaches the model.
Does this cover files, or only typed prompts?
Both, and for Claude Desktop that matters more than it does for a chat window. The agent works on files directly, so values inside documents are replaced at runtime the same way values typed into a prompt are.
Can we see what the agent actually did?
Yes. Top risks, actions taken, prompts processed and adoption by team, on one dashboard. Agents act on their own, so being able to see what was done after the fact is not optional.
How is Claude Desktop billed when it's secured through Mavs?
Claude Enterprise is licensed per seat. Through Mavs you pay for tokens consumed, billed against your Anthropic API licence, so cost follows usage rather than headcount. You can expand access to more users, while monitoring how they are using it via the productivity dashboard.
